Mind Over Memory

Aging, Recollection, and Environmental Support

An important clue to aging effects on episodic memory comes from findings that they are greatest when no specific cues are given (free recall), reduced when recall is cued, and often absent when the task is simply to recognize whether something has been previously encountered (Craik & McDowd, 1987). This has been explained as a failure of detailed recollection, as opposed to nonspecific recognition that something is familiar. Consistent with this, older adults are less likely to recall the jointly occurring features that make up events (associations) and their contexts, such as when and where they took place (Old & Naveh-Benjamin, 2008; Spencer & Raz, 1995). In contrast, familiarity is largely intact (Koen & Yonelinas, 2014). These difficulties may in part reflect basic problems in “binding” a unique memory trace due to neural damage in the hippocampus. However, failures of memory control resulting from prefrontal impairment are critical (Shing et al., 2010). This prompts the question of how control difficulties impact recollection. Converging evidence suggests that what happens prior to the point of retrieval is critical (Morcom & Rugg, 2004).

Craik (1983) proposed that older adults have difficulty initiating the necessary processes for recall but that, given adequate environmental support from external cues, their memory difficulties can be reduced or even abolished. The self-initiation impairment is attributed to a reduction in attentional resources. In line with this, reductions in recollection of detail and associations have been reproduced in young adults given a demanding secondary task to perform during a memory test. Other studies have manipulated environmental support. Stronger cues—for example, providing four rather than three letters of words to be recalled—can boost memory more in older adults than in younger adults (e.g., Angel et al., 2010), although not always (Park & Shaw, 1992). Encoding words with pictures, which supports distinctive processing, helps older adults more, as can external cues (Craik & Schloerscheidt, 2011). There is evidence too that when people try not to recall, external cues remain more potent reminders for older adults (Anderson, Reinholz, Kuhl, & Mayr, 2011). Nonetheless, while some environmental-support interventions do benefit older adults more, others are actually more helpful to the young—for example, instructions aimed at enhancing memory strategies. This variability may reflect the degree to which acting on the provided support itself depends on self-initiated processing (for review, see Luo & Craik, 2008).

Two relatively recent developments provide more traction on this important theoretical account. First, studies of attentional control have specified particular functions affected by aging. Braver’s dual mechanisms of control theory distinguishes proactive, anticipatory control from reactive control. Under proactive control, current goals are actively maintained by prefrontal cortex, allowing these goals—rather than external events—to drive behavior. This supports self-initiated behavior and minimizes interference from competing goals. In contrast, reactive control is transient and stimulus-driven and can help to resolve interference only once it emerges, potentially increasing overall control demands. An example of how this may apply to episodic memory is illustrated in Figure 1, explained further below. Older adults are impaired at proactive control in attentional tasks such as the AX Continuous Performance Test (AX-CPT), appearing to rely on reactive control (Bugg, 2014). When some information must be maintained in temporary memory (e.g., houses) and other information ignored (e.g., scenes), older adults are also less effective at selectively activating the goal-relevant information (Gazzaley, 2013).

OPEN IN VIEWER

Fig. 1.

Proactive and reactive control of episodic memory. In both panels, the goal is to recall people encountered in a particular context: last year’s holiday in Greece. The top panel illustrates proactive control, under which activation of this goal occurs before the external memory cues (pictures representing people) appear. This allows the relevant episode to be rapidly retrieved and interference from potentially competing episodes (e.g., a previous meeting with the man) prevented. The lower panel illustrates reactive control, under which proactive control is not engaged, so the retrieval goal is not activated until the cue pictures appear. The cues initially do not trigger full recollection, and because the second cue is also familiar, interference is also present from partial retrieval of another, goal-irrelevant episode (meeting him at a conference). This incurs a greater demand for monitoring. (The face images shown are drawn from the MUCT Landmarked Face Database; Milborrow, Morkel, & Nicolls, 2008.)

The second development involves brain imaging and behavioral methods designed to study preretrieval control (for review, see Mecklinger, 2010). Leading theories of memory control specify a role for proactive processes (Burgess & Shallice, 1996; Johnson & Raye, 2000). Effective memory cues reinstate contextual information that was incorporated in the memory trace and trigger retrieval when they match the memory trace (Tulving, 1983). They may remind us of where something took place or our thoughts at the time. But cues need not be external. People can strategically “self-cue” memory by internally generating cue information or elaborating on external cues in line with their goals. For example, visualizing a particular place before seeing a photo of a person may help us recall that we previously met him or her in that context (Fig. 1). This is one way in which people adopt a retrieval orientation reflecting specific retrieval goals. We may also engage a general retrieval mode—that is, an attentional set conducive to retrieval. Both are also referred to in terms of control operating at the “front end” to constrain what is retrieved (Jacoby, Shimizu, Velanova, & Rhodes, 2005).

However, these processes leading to recovery of a memory are difficult to study, since they cannot be separated from retrieval itself using measures of an ultimate memory judgment (Rugg & Wilding, 2000). This may be why the behavioral studies of self-initiation and environmental support discussed earlier have had mixed findings. Brain imaging has therefore become central to understanding how preretrieval control works. Event-related potentials (ERPs) reflecting scalp-recorded electrical activity and fMRI permit measures of how brain activity differs according to distinct retrieval goals, at different stages of retrieval. Using these approaches and novel behavioral measures, we, and others, have investigated the proposal that older adults do not fully engage control processes prior to retrieval (Morcom & Rugg, 2004).

Aging and Proactive Control of Memory

Four main experimental paradigms are used, illustrated in Figure 2. All involve an encoding (study) phase followed by a retrieval (test) phase. Items are studied in one of two alternative contexts (“deep” and “shallow” study tasks in the example in Fig. 2, left panel). This allows definition of different retrieval goals in the test phase in different blocks (or runs) of trials preceded by instructional cues. For example, participants are asked to target the deep task in a first run, then the shallow task, recognizing old items studied in the specified task (Fig. 2, center panel).

OPEN IN VIEWER

Fig. 2.

Experimental paradigms for the study of proactive control of episodic retrieval. All involve two main task phases in which items are studied in two or more alternative contexts and later retrieved separately from these different contexts. In the example, items (words) are committed to memory in the encoding (study) phase (left panel) while participants perform one of two orienting tasks: a “shallow” nonsemantic task or a “deep” semantic task (note that in paradigms 1 and 2, two semantic tasks are normally used in order to match retrieval difficulty, but the task combination shown is required for paradigm 3). In the initial retrieval (test) phase (center panel), the studied (old) words are re-presented along with unstudied new items (foils), and participants are asked to recognize studied items. In paradigm 1 (center panel), brain activity is measured in response to instructional cues that inform participants whether in upcoming trials they will be recognizing shallowly or deeply studied items. Paradigm 2 (center panel) measures how brain activity elicited by the unstudied retrieval cues (words) depends on the goal to recognize deeply or shallowly studied items (new items are also referred to as foils). In paradigm 3 (right panel), an additional, foil-memory test phase is used to assess the degree to which people recall more of the foils they first encountered when targeting deeply, as opposed to shallowly, studied information in the initial test phase (Jacoby, Shimizu, Velanova, & Rhodes, 2005). Paradigm 4 (center panel) assesses the downstream effects of retrieval goals on what is recollected. Neural activity associated with successful retrieval of targeted information (e.g., the word “PENGUIN” is a target when the goal is to recall items studied in the deep task) is compared for activity to new items; differences in activity are referred to as old/new effects. In this paradigm, some nontargets are also included to enable measures of goal-relevant (target) versus goal-irrelevant (nontarget) retrieval (Dywan, Segalowitz, & Webster, 1998). Nontargets are familiar items that were not studied in the currently targeted context (e.g., the word “BATH” was studied in the shallow task and appears as a nontarget while participants are targeting deeply studied items). See text for further details.

The first paradigm uses brain imaging to examine processes elicited by the instructional cues, before specific retrieval cues are even presented (Fig. 2, center panel). ERP studies in young adults have shown that preparatory neural activity depends on the type of information people have been asked to target in memory, indicating adoption of a retrieval orientation. This activity is goal-specific, distinct from preparatory activity that is common to different episodic retrieval tasks, suggesting engagement of a generic retrieval mode (Herron & Wilding, 2006). These ERP effects have not yet been studied in aging, but a recent fMRI study showed reduced preparatory activity in hippocampus during episodic retrieval in older adults and altered connectivity with prefrontal cortex. This was consistent with impaired proactive control—either retrieval orientation or mode—although age effects may not have been memory-specific (Dew, Buchler, Dobbins, & Cabeza, 2012). Given the substantial evidence that goal-directed preparatory attention is impaired in older adults in working memory and attention tasks (Gazzaley, 2013), this is likely to be a fruitful direction for future research.

The second paradigm also uses brain imaging. It examines the impact of retrieval orientation on how specific retrieval cues are processed (Rugg & Wilding, 2000). The critical comparison is of activity elicited by new (unstudied) items in the test phase under different retrieval goals (Fig. 2, center panel). Given that these items were not studied, this reflects the goals rather than retrieval itself (goal-related activity common to studied and new items can also be measured; Dobbins, Rice, Wagner, & Schacter, 2003). Given an established retrieval orientation, identical cues are processed differently according to what participants try to retrieve. In the example (Fig. 2), participants target words studied in either the deep or the shallow task. Alternative goals may relate to the format or location in which items were studied (see Mecklinger, 2010). Using this approach with ERPs, we found that older adults adopted a less distinctive retrieval orientation (Morcom & Rugg, 2004).

Converging evidence came from the third approach, the behavioral memory-for-foils paradigm (Jacoby et al., 2005). As in the second paradigm, analysis in this paradigm focuses on the new (unstudied) foils in the initial memory test (Fig. 2, center panel). A third task phase is now added to assess how these were processed under different retrieval goals (Fig. 2, right panel). It is thought that participants strategically boost recall in the initial retrieval phase by internally reinstating the targeted (deep or shallow) study context in response to the word cues. This cue elaboration determines the level of processing applied to the foils: Since deep foils are processed more meaningfully than shallow foils, memory is better for the deep foils in the final recognition test (Fig. 2, right panel). Unlike the first two (imaging) paradigms, which may detect any differential processing according to retrieval goal, this memory-for-foils effect depends on goal-specific reinstatement of the encoding context. Young, but not older, adults showed this effect.

A subsequent ERP study using the second paradigm examined age-related differences in retrieval orientation in a task requiring detailed recollection (Duverne, Motamedinia, & Rugg, 2009). In Morcom and Rugg’s (2004) and Jacoby et al.’s (2005) studies, initiation of recollection could be avoided if the simple “old” or “new” judgments were made on the basis of familiarity. Following Morcom and Rugg (2004), participants targeted items studied as either pictures or words, but in one condition, recall of source information (screen location at study) was also required. This time, older adults successfully adopted a retrieval orientation, but only when source recall was required. This is consistent with the idea of a self-initiation deficit. But it also suggests that older adults can, under some circumstances, effectively initiate goal-directed retrieval. It is currently unknown whether, once engaged, this control is always as robust. A key factor may be whether retrieval goals compete—that is, whether, unlike in Duverne et al.’s task, orientation to one type of to-be-retrieved information also requires rejection of items studied in other contexts or exclusion of other studied information. A recent fMRI study found that older adults’ prefrontal responses depended less than those of the young on which of two competing sources—study task or format—was targeted. Prefrontal responses were also less tightly coupled with posterior cortical regions carrying representations of the targeted information (Mitchell, Ankudowich, Durbin, Greene, & Johnson, 2013). This suggests impaired prefrontal goal maintenance, impeding ability to boost recall by reinstating study context in the posterior representation regions. However, although the data implicate goal-directed control, they do not speak to whether this control acted before or after the point of retrieval.

The first three paradigms are critical for separating preretrieval control from retrieval itself, but it is also important to establish the downstream effects of control on what is recollected. The fourth approach assesses neural responses associated with successful retrieval, typically comparing responses to remembered studied items and correctly rejected unstudied items (old/new effects; (Fig. 2, center panel). Dywan, Segalowitz, and Webster (1998) measured the degree to which the parietal ERP old/new effect, a neural correlate of recollection, favored targeted information (studied words) over nontargeted but familiar information (in this case, nontargets were novel words repeated during the test phase, but they may also be studied items from a nontargeted context; Fig. 2, center panel). Target and nontarget parietal effects were similar in magnitude in older adults and in young adults performing a demanding secondary task during retrieval, suggesting recollection was less selective. Similar studies in adolescents and young adults have shown that the degree of recollection of to-be-excluded information tracks individual and developmental differences in cognitive control and the speed with which retrieval orientation is engaged (Elward, Evans, & Wilding, 2013; Sprondel, Kipp, & Mecklinger, 2013)

Further examination of both preretrieval control and its downstream impact will be important because studies in young adults now point to more than one stage at which people can constrain what will be recollected (this work, like the studies of aging, has so far focused on goal-specific control—i.e., retrieval orientation rather than mode). Sometimes, it is possible to prevent recollection of nontargeted information entirely (gating; e.g., Morcom & Rugg, 2012). At other times, although there is irrelevant recollection, this is constrained to some degree by the retrieval orientation (McDuff, Frankel, & Norman, 2009). This may indicate a second, later stage of constraint involving direct goal-dependent modulation of the neural reactivation during recollection (Elward & Rugg, 2015). A key question for aging research will be which of these stages is affected, and which preserved. Future studies should clarify how existing indices of preretrieval control relate to constraints on recollection, and it would be helpful to combine the different approaches. More definitive links are also needed between proactive control as measured in attentional paradigms such as the AX-CPT and episodic preretrieval control. Elward and Wilding’s work in young adults has provided initial evidence that retrieval selectivity tracks individual and experimental variations in cognitive control.

Proactive Control, Reactive Control, and Monitoring

The suggestion that aging impairs proactive control does not, of course, mean that this is its only effect on episodic memory. Control difficulties likely also impact encoding (e.g., Old & Naveh-Benjamin, 2008) and what occurs after information is retrieved. The source-monitoring framework outlines reflective, self-generated processes required to specify the source, or context, of our mental experiences (Johnson & Raye, 2000). Proactive cue-elaboration processes are part of the framework, but a central focus is on monitoring the products of recall to prevent memory errors. Some of these reactive processes also depend on prefrontal cortex and are altered in aging (for review, see Gallo, 2006). Behavioral data indicate that older adults are less likely to engage specific monitoring strategies, such as recall-to-reject (“I know I didn’t see this item, because I remember hearing it”), to avoid memory errors. Susceptibility to errors is also greater in individuals with poorer prefrontal function as measured using standardized cognitive tests.

One possibility is that proactive and reactive memory control both suffer as a result of prefrontal cortex decline. However, impaired proactive control may also increase demands on reactive control (Braver & West, 2008). At least in attentional tasks, reactive control may even be spared (Bugg, 2014). The load-shift hypothesis proposes that impaired recollection primarily reflects preretrieval difficulties, with a consequent greater shift to, and load on, postretrieval monitoring (Velanova, Lustig, Jacoby, & Buckner, 2007). Consistent with this, older adults show increased neural activity during postretrieval monitoring, which is less goal-specific (McDonough, Wong, & Gallo, 2013). If this proposal is correct, alterations in postretrieval monitoring and evaluation may reflect a form of compensation for proactive failure (for a similar view, see Dew et al., 2012). The degree to which preretrieval and postretrieval impairments coexist or trade off against one another is a key question for future research. It should also inform interpretation of findings of additional brain activity in older adults in imaging studies (Morcom & Johnson, 2015).

Conclusions

Current data support a prima facie case that initiation of episodic memory retrieval is impaired in aging, reducing ability to direct retrieval according to goals (see also Wilkens, Erickson, & Wheeler, 2012). But the details remain to be established. To borrow an expression from Light (1991), there are three main (nonexclusive) “hypotheses in search of data.” The first is that initiation of proactive control is impaired, but, once initiated, the control is effective (Duverne et al., 2009; Morcom & Rugg, 2004). The second assumes that establishment of control is impaired only in the presence of interference, consistent with findings from studies of temporary memory and attention (Gazzaley, 2013). The third proposes that control can be established but is less effective in terms of downstream constraint of what is retrieved (Dywan et al., 1998).

These questions bear critically on how older adults should best respond to memory difficulties. In a recent review, Lindenberger and Mayr (2014) suggested that the environmental-support account can explain a range of age-related differences in perceptual and attentional function as well as memory. Pinpointing the nature of memory difficulties is important not only for understanding them but also for training and remediation: Should reliance on the kinds of inner control strategy optimal in youth be encouraged, along with training aimed at bolstering these abilities, or are aging brains best advised to lean more heavily on external cuing and support? As life spans increase and technologies improve, these decisions will be critical.